Cold Plasma-Induced Modification of the Adsorption Characteristics of Carrot: Physics-Constrained Neural Network Modelling, Adsorption Isotherms and Thermodynamic Insights

Abstract

Carrot slices were treated on both sides with an argon-propagated cold plasma jet for 30, 60 and 90s and tagged CPJ-30, CPJ-60 and CPJ-90, respectively. The samples were dried in a vacuum dryer and tray dryer at 70 °C and their equilibrium moisture content (EMC) was evaluated gravimetrically. Sorption isotherms were analysed using classical adsorption models. A new model was developed, called Logarithmic-Quadratic Moisture Sorption (LQMS). The EMC was predicted using a physics-constrained neural network (PCNN) constrained with the Guggenheim-Anderson-de Boer (GAB) model and an unconstrained neural network. The thermodynamic properties were determined along with FTIR analysis. Cold plasma treatment modified the accessibility and strength of water-binding sites. In vacuum-dried samples, the isosteric heat decreased from 212.96-24,720.26 J/mol in the control to 116.20–8879.46 J/mol in the treated carrots. Similarly, the isosteric heat for tray-dried carrots decreased from 1242.54-13,152.75 J/mol in the control sample to 1703 − 706,948.92 J/mol in the treated carrots. The reduction in O-H-associated FTIR intensity supports changes in the EMC. The entropy of vacuum-dried samples decreased from 148.78 to 199.86 J/mol.K in the control sample to 144.49–181.05 J/mol.K, in the treated carrots. The entropy of tray-dried samples also decreased from 45.29 to 73.69 J/mol.K in control sample to 27.10–62.34 J/mol.K, in the treated carrots. The LQMS model showed improved performance against its counterpart models and gave a higher validation performance. The PCNN model gave improved validation performance against GAB model and neural network model. These findings provide useful information for packaging design and storage conditions for dried carrots.